Hi, do you think the category "Fuzzing Machine Learning Model" is too broad and misleading?

[haoyang9804]

I think we can remain the following papers in this category

• MDPFuzz: Testing Models Solving Markov Decision Processes
• RapidFuzz: Accelerating fuzzing via Generative Adversarial Networks
• CoCoFuzzing: Testing Neural Code Models with Coverage-Guided
• Fuzz Testing based Data Augmentation to Improve Robustness of Deep Neural Networks
• Coverage Guided Differential Adversarial Testing of Deep Learning Systems
• CAGFuzz: Coverage-Guided Adversarial Generative Fuzzing Testing of Deep Learning Systems
• DeepHunter: A Coverage-Guided Fuzz Testing Framework for Deep Neural Networks
• TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing
• DLFuzz: Differential Fuzzing Testing of Deep Learning Systems

As for others, except for Graph-based Fuzz Testing for Deep Learning Inference Engines, they belong to another category possibly titled "Fuzzing Deep Learning Libraries".

As for Graph-based Fuzz Testing for Deep Learning Inference Engines, a deep learning inference engine is a compiler-like tool (not a compiler but contains compilation). I'd like to categorize it into another new category, "Fuzzing Deep Learning Compilers". In this category, there are also several other works, such as:

• Fuzzing Deep Learning Compilers with HirGen [ISSTA '23]
• NNSmith: Generating Diverse and Valid Test Cases for Deep Learning Compilers [ASPLOS '23]

Do you think it's a good idea to re-organize these papers? After the reorganization, I'd like to add some papers belonging to "Fuzzing Deep Learning Compilers" category.

[wcventure]

Yes, I quite agree with you about that. I think the category "Fuzzing Machine Learning Model" is too broad. Can you create a pull request according to your idea?

[haoyang9804]

Yes, I quite agree with you about that. I think the category "Fuzzing Machine Learning Model" is too broad. Can you create a pull request according to your idea?

Hi, I have create a pull request. Please take a look. To be honest, I'm not an expert in fuzzing deep learning libraries so the paper listed may not be complete. I think we can also create a new subsection titled compiler fuzzing. But since I have created one here, I think you can put this link about maintaining compiler correctness (including compiler fuzzing) in the README if you like.

[wcventure]

https://haoyang9804.github.io/CompilerCorrectness/ Great!